Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Dual-Layer Detector Cone-Beam CT Angiography for Stroke Assessment: First-in-Human Results (the Next Generation X-ray Imaging System Trial)

Dual-Layer Detector Cone-Beam CT Angiography for Stroke Assessment: First-in-Human Results (the... Dual-Layer Detector Cone-Beam CT Angiography for Stroke Assessment: First-in-Human Results (the Next Generation X-ray Imaging System Trial) F. Ståhl, H. Almqvist, J. Kolloch, Å. Aspelin, V. Gontu, E. Hummel, M. van Vlimmeren, M. Simon, A. Thran, Å. This information is current as Holmberg, M.V. Mazya, M. Söderman and A.F. Delgado of June 3, 2023. AJNR Am J Neuroradiol 2023, 44 (5) 523-529 doi: https://doi.org/10.3174/ajnr.A7835 http://www.ajnr.org/content/44/5/523 ORIGINAL RESEARCH ADULT BRAIN Dual-Layer Detector Cone-Beam CT Angiography for Stroke Assessment: First-in-Human Results (the Next Generation X-ray Imaging System Trial) F. Ståhl, H. Almqvist, J. Kolloch, Å. Aspelin, V. Gontu, E. Hummel, M. van Vlimmeren, M. Simon, A. Thran, Å. Holmberg, M.V. Mazya, M. Söderman, and A.F. Delgado ABSTRACT BACKGROUND AND PURPOSE: In patients with stroke, IV cone-beam CTA in the angiography suite could be an alternative to CTA to shorten the door-to-thrombectomy time. However, image quality in cone-beam CTA is typically limited by artifacts. This study evaluated a prototype dual-layer detector cone-beam CT angiography versus CTA in patients with stroke. MATERIALS AND METHODS: A prospective, single-center trial enrolled consecutive patients with ischemic or hemorrhagic stroke on initial CT. Intracranial arterial segment vessel conspicuity and artifact presence were evaluated on dual-layer cone-beam CTA 70-keV virtual monoenergetic images and CTA. Eleven predetermined vessel segments were matched for every patient. Twelve patients were necessary to show noninferiority to CTA. Noninferiority was determined by the exact binomial test; the 1-sided lower performance boundary was prospectively set to 80% (98.75% CI). RESULTS: Twenty-one patients had matched image sets (mean age, 72 years). After excluding examinations with movement or con- trast media injection issues, all readers individually considered dual-layer cone-beam CT angiography noninferior to CTA (CI bound- ary, 93%, 84%, 80%, respectively) when evaluating arteries relevant in candidates for intracranial thrombectomy. Artifacts were more prevalent compared with CTA. The majority assessment rated each individual segment except M1 as having noninferior con- spicuity compared with CTA. CONCLUSIONS: In a single-center stroke setting, dual-layer detector cone-beam CTA virtual monoenergetic images are noninferior to CTA under certain conditions. Notably, the prototype is hampered by a long scan time and is not capable of contrast media bolus tracking. After excluding examinations with such scan issues, readers considered dual-layer detector cone-beam CTA noninfe- rior to CTA, despite more artifacts. ABBREVIATIONS: CBCT ¼ cone-beam CT; CBCTA ¼ cone-beam CTA; CNR ¼ contrast-to-noise ratio; DL ¼ dual-layer; VMI ¼ virtual monoenergetic images one-beam CT (CBCT) with flat image detectors was first (CBCTA) in the angiography suite could be an alternative to 1 6 Cdemonstrated in 2000. The technique is widely used in CTA to shorten the door-to-thrombectomy time. CBCTA has interventional radiology procedures for anatomic and pathology been shown to be equal or better than CTA for the diagnosis of 7,8 2-5 intracranial stenosis or proximal MCA occlusions. Astudy in assessment. In patients with stroke, IV CBCT angiography 10 patients indicated superior image-quality results in several CBCTA arterial segments compared with CTA; however, the Received October 13, 2022; accepted after revision February 27, 2023. patient cohort was heterogeneous and biased, and the results From the Departments of Neuroradiology (F.S., H.A., J.K., Å.A., V.G., Å.H., M. Söderman, A.F.D.) and Neurology (M.V.M.), Karolinska University Hospital, Stockholm, Sweden; were only presented for individual vessel segments. Aretrospec- Department of Clinical Neuroscience (F.S., H.A., V.G., M.V.M., M. Söderman, A.F.D.) tive study on 16 patients with stroke with large-vessel occlusion Karolinska Institutet, Stockholm, Sweden; Image Guided Therapy (E.H., M.v.V.), Phillips Healthcare, Best, the Netherlands; and Philips Research Hamburg (M. Simon, A.T.), of the anterior circulation indicated that CBCTA generated from Hamburg, Germany. a volume perfusion scan can reliably identify the site of occlusion The trial was supported by a grant from the European Commission (Horizon 2020, in the ICA and M1 segment of the MCA. However, this study NEXIS-project, grant number 780026). did not find CBCTA superior to CTA. Please address correspondence to Fredrik Ståhl, MD, Department of Neuroradiology, Karolinska University Hospital, Eugeniavaegen 3, 171 76 Stockholm, Sweden; e-mail: In CTA, virtual monoenergetic images (VMIs) reconstructed Fredrik.stahl@regionstockholm.se from dual-energy scans have superior image quality and improve Indicates open access to non-subscribers at www.ajnr.org the diagnostic assessment of intracranial vessels compared with 11-13 Indicates article with online supplemental data. conventional polyenergetic images. CBCTA image quality would http://dx.doi.org/10.3174/ajnr.A7835 likely benefit from dual-energy VMI reconstructions by means of AJNR Am J Neuroradiol 44:523–29 May 2023 www.ajnr.org 523 reduced beam-hardening artifacts and an improved contrast-to- the dual-layer detector captures more photons than a conventional 11-14 noise ratio (CNR). Recently, performance characteristics of a CBCT detector. DL-CBCTA VMIs were derived from Compton and photoelectric base projections (basis material decomposition prototype dual-layer detector CBCT (DL-CBCT) system were was performed in the projection domain). Vessel conspicuity may published, with the possibility of optimizing CNR by VMI energy be improved due to intrinsic compensation of iodine beam-harden- selection. ing artifacts in DL-CBCTA VMIs. Anticorrelated noise reduction Because previous studies on CBCTA are small, heterogeneous, was exploited as an inherent benefit of DL-CBCT. All DL-CBCTA and selective, a trial enrolling consecutive patients is necessary to and CTA scans used the same automated injection protocol: 85 mL determine the role of CBCTA in the primary diagnostic setting of 7-10 of iodine, 320 mg I/mL (iodixanol, Visipaque; GE Healthcare), suspected stroke. In this prospective trial, we assessed arterial 5 mL/s into a peripheral vein followed by an 80-mL saline chaser. visibility and artifact presence in patients with stroke when exam- The second-phase CTA standard reconstruction was included ined with DL-CBCT angiography (DL-CBCTA) and CTA. To our in the study as a reference standard because it was reconstructed knowledge, this is the first trial on CBCTA with prospectively with a smaller FOV and the timing was typically more similar to defined end points, enrolling awake patients with the aim of inves- the DL-CBCTA contrast phase (late arterial/arteriovenous). tigating noninferiority to CTA. Furthermore, this work presents Phantom measurements indicated that the DL-CBCTA and CTA the first-in-human results from a DL-CBCT prototype system. systems had comparable high-contrast spatial resolution. Scan details are shown in Table 1. In conjunction with the CTA and MATERIALS AND METHODS DL-CBCTA, the quality of the image acquisition was evaluated Trial Design and Participants with regard to patient motion and timing of the contrast media The Next Generation X-ray Imaging System (NEXIS) trial was a injection. Scans were rated as good, acceptable, or poor. prospective, nonrandomized open-label single-sequence, 2-period crossover clinical trial with blinded image readers conducted at a Diagnostic Image-Quality Assessment comprehensive stroke center. The study sought to include patients A series of pilot studies were performed to determine the optimal evaluated for endovascular thrombectomy on arrival at the hospi- VMI energy and noise-reduction levels for the DL-CBCTA tal. Patients with a suspected large-vessel occlusion of the anterior images. We randomized 50- to 80-keV VMIs at 3 different noise- circulation underwent prehospital triage to bypass primary stroke 16 reduction levels, and expert readers evaluated them side-by-side centers as per clinical routine. Patients 50 years of age or older blinded to noise and energy levels (F.S., H.A., V.G., M. Söderman, with ischemic stroke of the anterior circulation or hemorrhagic A.F.D.). Window settings were normalized to a previous publica- stroke were consecutively enrolled from November 2020 to April tion on dual-energy CTA. Images were ranked in order of pref- 2021. Initial imaging was with CTA as per clinical routine and sub- erence with regard to image quality and artifact presence. The 70- sequently with the prototype DL-CBCT in an adjacent room. keV images with a moderate noise reduction were best suited for Depending on applicable eligibility criteria, patients were imaged the purpose of the study. once or twice with DL-CBCT (the same day as CTA or/and 1 day af- In the reader study, vessel conspicuity and artifact presence ter CTA). Eligibility criteria, inclusion groups, and flow diagrams of were evaluated separately on 5-point Likert scales, adopted with study participation arepresented in theOnlineSupplemental Data. slight modifications from previous studies (5, excellent vessel con- The study protocol and informed consent forms were approved spicuity or no artifacts; 1, vessel not visible or extensive artifacts) by the Swedish Ethical Review Authority (approval No. 2020– 10,19 (see the Online Supplemental Data for a detailed description). 00157). The prototype DL-CBCT system was approved by the All patients had unilateral ischemic or hemorrhagic stroke; thus, Swedish Medical Products Agency (document No. 5.1–2020– the affected hemispheres were not included in the analysis. Images 6325), in accordance with the Medical Device Directive. Philips were randomized and evaluated independently by 3 neuroradiolo- Healthcare was the formal sponsor of the study because of direct gists (H.A., J.K., and Å.A.) each with .9 years of experience in a liability for the prototype DL-CBCT system. An independent quali- single-sequence, 2-period crossover design with a mean washout fied research organization contracted by the sponsor monitored the period of 4 weeks (range, 1–8 weeks). The readers were blinded to study. All patients signed an informed consent. The trial was sup- technique, and there was no clinical information. The study soft- ported by a grant from the European Commission (Horizon 2020, ware allowed changes in section thickness, viewing plane, window NEXIS-project, grant No. 780026). The study was registered pro- level, and window width. spectively at clinicaltrials.gov (identifier: NCT04571099). Sixteen intracranial arterial segments were prospectively defined for the readers’ study. For the powered analysis, some CTA and Prototype DL-CBCTA arterial segments were merged to render 11 arterial segments The systems used for CTA were Aquilion ONE (Canon Medical Systems) and IQon Spectral CT (Philips Healthcare). The examina- per patient (Table 2). For merged segments, the scores of indi- tion was according to clinical routine, typically including noncon- vidually evaluated segments were averaged. The score difference trast CT of the head, multiphase CTA, and CTP. The prototype between DL-CBCTA and CTA for each segment in each patient DL-CBCT system (Allura NEXIS Investigational Device; Philips determined whether a segment was considered inferior, equal, Healthcare) was a commercial interventional C-arm x-ray system or superior. In case of an absent segment due to variant anat- omy, a vessel conspicuity score of 1 was given for the specific (Allura Xper FD20/15; Philips Healthcare) fitted with a dual-layer segment (see the previous paragraph). In addition to individual 20-inch (379.4  293.2 mm) non-CE marked detector prototype. The prototype detector has been previously described. In essence, reader’s results, the majority assessment is also presented, ie, 524 Ståhl May 2023 www.ajnr.org Table 1: Scan details Canon Aquilion ONE Philips IQon Prototype DL-CBCT Tube (kV) 100 120 120 Tube current (mAs/mA) (average) 196 (Auto-modulation) 114 (Auto-modulation) 310 Scan time/rotation time (sec) 3.3/0.5 (Full rotation) 2.5/0.3 (Full rotation) 20.0/20.0 (200° rotation) Nominal beam width (mm) 80  0.500 64  0.625 194.700 Pitch factor 0.813 0.671 NA Display FOV coronal  sagittal  axial (mm ) 210.9  210.9  160.0 210.0  210.0  160.0 251.8  251.8  194.7 Section thickness (mm) 0.50 0.67 0.66 Matrix size 512  512 512  512 384  384 15a Reconstruction kernel FC43 Filter UA Ståhl et al 4 15a Reconstruction algorithm AIDR 3D eStandard iDose level 4 Ståhl et al Avg CTDIvol (16-cm phantom) 20.0 mGy 21.3 mGy NA Air kerma (in an 18-cm water phantom) NA NA 57.6 mGy MTF 50%: 3.78 50%: 3.46 50%: 3.57 10%: 6.57 10%: 6.65 10%: 6.04 Note:—CTDIvol indicates volume CT dose index; MTF, modulation transfer function; NA, not applicable. a 15 Details of the prototype algorithm are described in Ståhl et al. Air kerma in an 18-cm diameter plastic water phantom at the center of the scan length, measured in accordance with American Association of Physicists in Medicine Task Group Report 111. Generated from consecutive scans of the upper bead of a Catphan CTP528 module (The Phantom Laboratory). Table 2: Arterial segments RESULTS 11 Segments 11 Segments Population Characteristics 16 Segments (Powered) (Thrombectomy) Of 28 consecutively enrolled patients, 5 had no in-house CTA ICA ICA ICA and 2 underwent subcutaneous IV contrast media injection dur- M1 M1 M1 ing the DL-CBCTA scan. Two patients were imaged twice with M2 M2 M2 DL-CBCTA, and for those patients, the results from both scans M3 M3-M4 M3 were averaged. Consequently, 21 complete and matched DL- M4 M4 A1 A1-A2 A1 CBCTA and CTA image sets from 21 patients were included A2 A2 (Fig 1, flow diagram). The mean age was 72 (SD, 9) years, and Lenticulostriate Lenticulostriate 14 were women (67%). The right hemisphere was affected in 11 Vertebral Vertebral Vertebral (57%) patients. Three patients had a hemorrhagic stroke. Of the Basilar Basilar Basilar AICA AICA-PICA-SCA remaining 18 patients with ischemic stroke, 17 (94%) presented PICA with occlusion of the ICA or proximal MCA (M1 or M2 seg- SCA ment) on CTA. Seventeen patients (81%) were imaged with DL- Basilar perforators Basilar perforators CBCTA the day after CTA imaging (mean, 23.8 [SD, 3.0] hours), P1 P1-P2 P1 of whom 15 had been treated with thrombectomy the day before P2 P2 (modified TICI grade 2C reperfusion or better in 11 [73%]). Note:—Lenticulostriate indicates lenticulostriate artery perforators; Vertebral, in- tracranial vertebral artery; Basilar perforators, basilar artery perforating branches; Because only the unaffected hemisphere was included in the sta- SCA, superior cerebellar artery. tistical analysis, the degree of reperfusion was not considered to influence the results. Four patients (19%) were only imaged segments judged as superior/equal or inferior compared to CTA with DL-CBCTA on the same day as CTA (mean, 1.2 [SD, 0.6] by at least 2 out of 3 readers. hours after CTA). Statistical Analysis The powered outcome of this study was the proportion of arterial Vessel Visibility Assessment segments with equal or superior visibility compared with CTA For the powered end point of overall arterial visibility, 21 patients (Table 2). The target proportion was 90% equal or superior ratings, with a total of 231 matched arterial segments (11 per patient) with a lower boundary performance goal set to 80%. The 1-sided a were evaluated by each reader. One reader scored equal or supe- was modified from .025 to .0125, to account for multiple end rior vessel visibility for DL-CBCTA and CTA in 90% of arterial points. A sample size of 126 was required to render 80% power. segments (CI lower boundary of 84%) and met the predefined Because the powered end point was trait-based, 12 patients with a total of 132 arterial segments (11 segments each) constituted the noninferiority criteria (lower performance boundary of .80%), minimum sample size. For these analyses, the exact binomial test whereas the other 2 readers did not (CI lower performance with a 1-sided 98.75% CI was used. Within-subject correlation of boundary of 58% and 53%, respectively). The result of the major- arterial segments was assessed by the Cochran-Mantel-Haenszel ity analysis was 77% of arterial segments being rated noninferior 20 21 test. Interrater agreement was assessed by the Fleiss k. in DL-CBCTA (CI lower boundary of 71%). The Fleiss k between readers was 0.25 (fair agreement). Examples of patient scans with Statistical analyses and graphic presentations were made in R no disagreement among readers and with considerable interrater Studio (Version 1.4.1103; http://rstudio.org/download/desktop). A biostatistician was consulted at all stages of the study. variability are presented in the Online Supplemental Data. AJNR Am J Neuroradiol 44:523–29 May 2023 www.ajnr.org 525 98%, 91%, and 88%, and all readers individually exceeded the CI Post Hoc Subset Analyses lower performance boundary of 80% (Table 3). Majority results Nine of the 21 (43%) DL-CBCTA scans were of poor quality due were identical to those of the powered subset (98%, CI lower bound- to patient motion or suboptimal timing of the contrast media ary 93%). The Fleiss k for this data set was 0.38 (fair agreement). injection (Fig 2). No CTA scan was rated as poor. In a subset There was no significant within-patient correlation of individual ar- analysis including only the 12 matched scans of acceptable or good quality, the 3 readers rated vessel visibility as noninferior terial segments for any of the individual readers in the thrombec- in 98%, 88%, and 78% of the cases, respectively, which meant tomy data set or subset. However, a significant correlation was that 2 readers met the predefined performance goal (CI lower found for all majority analyses and most individual reader results boundary of .80%). The majority result for this subset was 98% for the powered data set and subset (Online Supplemental Data). (CI lower boundary of 93%). See Table 3 for a detailed display of the results. Image Artifact Assessment Eleven arterial segments especially relevant for thrombectomy Image artifacts were assessed in the same arterial segments as used candidates were selected for evaluation in the same patient subset for vessel visibility. For the prospectively determined arterial seg- (Table 2). According to each reader, noninferior vessel visibility was ments, readers perceived artifact presence as noninferior to CTA in 58%, 42%, and 36% of the segments rated (Table 4). The Fleiss k was 0.26 (fair agreement). After excluding the 9 scans of inferior quality, noninferiority to CTA for artifact assessment was per- ceived in 68%, 66%, and 55% of the cases. The lowest number of artifacts compared with CTA was observed in the subset of 11 arterial segments espe- cially relevant for thrombectomy candi- dates (Table 4). The Fleiss k for this data set was 0.34 (fair agreement). In summary, noninferiority in terms of artifacts was not seen in the full data set or any subset analysis, given a lower performance boundary of 80%. Evaluation of Individual Arterial FIG 1. Of 28 consecutively enrolled patients, 5 had no in-house CTA and 2 had subcutaneous IV Segments contrast media injection during the DL-CBCTA scan. Two patients were imaged twice with DL- CBCTA, and for those, the results from both scans were averaged. Consequently, 21 complete The Online Supplemental Data show and matched DL-CBCTA and CTA image sets from 21 patients were included. the whole data set of individual arterial FIG 2. DL-CBCTA 70-keV images (upper row)and CTA (lower row) with MIP 35-mm section thickness. A and B, Acceptable-quality DL-CBCTA scans. C and D, Typical scans in the data set affected by motion artifacts. Lower row (E–H) shows the corresponding CTA. Note that images F, G, and H show a right-sided MCA occlusion, which have been resolved at the time for DL-CBCTA imaging. Only the arterial anatomy in the unaf- fected hemisphere was evaluated in this study. 526 Ståhl May 2023 www.ajnr.org a Table 3: Vessel visibility DISCUSSION Powered Data Powered Thrombectomy Data Thrombectomy Thestudy compared theintracranial ar- Set Subset Set Subset tery visibility of IV DL- CBCTA images Patients 21 12 21 12 with the reference standard CTA in con- Segments 231 132 231 132 secutively enrolled patients with stroke. rated b b Of dual-detector CBCTA scans, 43% Majority 0.77 (0.70) 0.98 (0.93) 0.77 (0.71) 0.98 (0.93) b b were of poor quality due to movement Reader 1 0.65 (0.58) 0.88 (0.80) 0.67 (0.60) 0.91 (0.84) b b b b Reader 2 0.90 (0.84) 0.98 (0.93) 0.89 (0.83) 0.98 (0.93) artifacts or contrast media injection Reader 3 0.60 (0.53) 0.78 (0.69) 0.68 (0.61) 0.88 (0.80) issues. In the subset of patients with ac- Proportion of DL-CBCTA arterial segment visibility rated equal or superior to CTA. The data set (21 patients) ceptable scan quality, DL-CBCTA was includes all scans; the subset (12 patients) excluded inferior scans. The 98.75% CI of the 1-sided lower performance noninferior to CTA with regard to intra- boundary is in parentheses (lower boundary is defined as 80% rated equal or superior). Statistically significant result. cranial artery conspicuity, despite more prominent image artifacts. The strengths of the trial were the Table 4: Artifacts prospectively defined end points, the Powered Data Powered Thrombectomy Data Thrombectomy clinically relevant study population, Set Subset Set Subset and the assessment by blinded readers. Patients 21 12 21 12 Segments 231 132 231 132 The noninferiority lower boundary was rated set to 80% to minimize the risk of type Majority 0.41 (0.34) 0.63 (0.53) 0.55 (0.48) 0.81 (0.72) I error. For the complete data set, 1 Reader 1 0.42 (0.35) 0.68 (0.58) 0.54 (0.46) 0.85 (0.77) reader determined DL-CBCTA arterial Reader 2 0.58 (0.50) 0.66 (0.56) 0.65 (0.58) 0.74 (0.64) visibility as noninferior to CTA, but the Reader 3 0.36 (0.29) 0.55 (0.44) 0.52 (0.44) 0.73 (0.63) a other 2 readers did not. However, after Proportion of DL-CBCTA arterial segment artifacts rated equal or superior to CTA. Data set (21 patients) includes all scans; subset (12 patients) excluded inferior scans. The 98.75% CI of the 1-sided lower performance boundary is excluding DL-CBCTA scans with in parentheses (lower boundary is defined as 80% rated equal or superior). movement artifacts or suboptimal con- trast media injections, 2 readers indi- vidually agreed on noninferiority of DL-CBCTA versus CTA. In this subset, the majority score deter- segment ratings per reader and according to the majority assess- mined vessel conspicuity to be noninferior to CTA. Image arti- ment. In summary,A2-,M4- and AICA-segment vesselcon- facts were generally more abundant in the DL-CBCTA images, spicuity was most commonly rated as noninferior to CTA (at and the results did not reach noninferiority. least 76% for each reader and segment), with high majority After we excluded scans of poor quality, all arterial segments scores (range, 81%–90% rated noninferior). The lenticulostri- except M1 showed noninferior conspicuity to CTA according to ate and basilar perforating arteries had high majority scores the majority assessment. With regard to individual reader results, for vessel conspicuity; however, 1 reader’s results largely dif- theM4, A2, P2, and AICA segments had thebestvessel visibility in fered from the other 2. It was noted that a safe distinction the subset, and all except AICA expectedly indicated a low degree from small draining veins was difficult. Consequently, the ma- of artifacts in the corresponding segments. The M1 and ICA seg- jority score was interpreted with caution for these segments. ments had the lowest scores for vessel visibility compared with The A2 segment was rated with the fewest artifacts (majority CTA and indicated a higher prevalence of artifacts. Our results score, 76% rated noninferior), followed by A1, M2, M3, M4, indicate that arteries that are not influenced by skull base artifacts P1, and P2 segments (majority score, 62%–67%, were rated may have superior conspicuity on DL-CBCTA compared with noninferior). CTA. As the systems had comparable spatial resolution, this may In the subset of 12 acceptable scans (Online Supplemental be attributed to fundamental differences of the VMI (DL-CBCTA) Data), A2, M4, AICA, and P2 segments were the most commonly and polyenergetic (CTA) reconstructions. DL-CBCTA VMIs had a rated as noninferior vessel conspicuity to CTA (at least 92% for lower absolute noise (achieved through anticorrelated noise reduc- each reader and segment). According to the majority assessment, tion) and may be superior in mitigating beam-hardening artifacts 100% noninferior vessel conspicuity to CTA was seen for all seg- from iodinated vessels (due to material decomposition in the pro- ments except M1 (75% rated noninferior). For artifacts, 100% of jection space) compared with CTA. In contrast, arteries adjacent to A1, A2, M3, M4, and P2 segments were noninferior in the subset the skull base showed an inferior visibility compared with CTA. according to the majority assessment. Even though beam-hardening artifacts are mitigated by VMIs, the M1-segment vessel conspicuity was noninferior to CTA in findings were expected because circular CBCT scans are highly 57% of the whole data set and 75% of the subset, the lowest susceptible to skull base artifacts such as beam-hardening. Since scores registered. The M1 segment also had a higher degree of the M1 and ICA segments are often involved in acute ischemic artifacts (43% rated noninferior in the data set and 67% in the stroke, future effort should focus on improving visualization of subset). The intracranial ICA segment similarly had a high these segments. Recently, a novel x-ray tube trajectory for CBCT prevalence of artifacts, and its scores were among the lowest was shown to improve intracranial image quality by reducing bone registered. beam-hardening artifacts from the skull. AJNR Am J Neuroradiol 44:523–29 May 2023 www.ajnr.org 527 The DL-CBCTA scan time was 20 seconds to enable dual- A minority of patients (19%) were imaged with DL-CBCTA layer image acquisition. Nine scans (43%) were considered of shortly after CTA, which may have affected the distribution of io- poor quality, predominantly due to motion artifacts. No repeat dine contrast media in the DL-CBCTA images. The impact of scans were performed in this study, in accordance with the potential differences in flow dynamics and vessel status for patients ethics approval and study protocol. Most interesting, all except imaged with DL-CBCTA 1 day after CTA is unknown. CNR com- 1 DL-CBCTA scan obtained on the day of admission showed parison of DL-CBCTA and CTA was not possible because image some degree of motion artifacts despite efforts to fixate the head acquisition timing after IV contrast injections were not identical. A in the head rest. The long scan time is a characteristic disadvant- previous study on CBCTA used a 512  512 image pixel matrix, age of CBCT, which must be considered when selecting an similar to CT. Using a 384  384 image pixel matrix for DL- appropriate imaging technique for patients with stroke. In CBCTA impacts the level of detail and noise characteristics, possi- bly affecting the perception of image quality. The DL-CBCTA scan awake patients with limited compliance, efforts to minimize the time of 20 seconds made it prone to movement artifacts. Also, in risk of patient movement is warranted. An enhanced head fixa- some cases, image quality was poor due to suboptimal timing of tion may decrease motion artifacts. Moreover, novel recon- the contrast media injection. Because the trial studied a prototype struction methods may mitigate the effects of patient motion in 24,25 system,manualbolus tracking wasnot possible.Instead,a manual CBCT. If there are still considerable motion artifacts pres- delay of 15–20 seconds was used for all patients. Furthermore, we ent, one may consider a repeat scan. used only 1 acquisition protocol. Future studies should involve Earlier studies evaluating the diagnostic accuracy of CBCTA image acquisition and reconstruction protocol optimization. either had patients under general anesthesia or a highly selective DL-CBCTA after IV contrast media injection in angiography patient cohort able to comply with instructions. As expected, suite enables the primary diagnostic work-up and subsequent treat- these studies did not report any significant impact on image qual- ment in the same room. In the clinical setting, this approach would ity from motion artifacts. In 1 study, it is unclear whether analy- require concurrent image evaluation of the brain to rule out intra- ses were matched pair-wise, and the results were presented only cranial hemorrhage and to enable assessment of irreversible brain is- for individual vessel segments and not for the CBCTA as a chemia and possibly brain perfusion. Noncontrast DL-CBCT of the whole. In addition, outcome measures from previous studies brain has been studied within the scope of the NEXIS trial, and data were not prospectively defined and typically were not adjusted 7-10 are currently being prepared for final analysis. Visualization of the for multiple comparisons. cervical vasculature is warranted in patients eligible for thrombec- This trial has several limitations and technical considerations. tomy; however, this was not addressed in the scope of this study. The ethics approval did not allow inclusion of patients before treat- ment. Thus, nonsymptomatic vessel segment conspicuity and arti- facts were evaluated as characteristics of diagnostic quality. The CONCLUSIONS original intention was to perform a subset analysis on the diagnos- In a single-center stroke setting, DL-CBCTA 70-keV virtual tic accuracy to identify intracranial occlusions in a subgroup of monoenergetic images are noninferior to CTA under certain con- patients (inclusion group 3.I, Online Supplemental Data), but the ditions. Notably, the prototype system is hampered by a long analysis was not performed due to few patients (n ¼ 3, of which 2 scan time and is not capable of contrast media bolus tracking. After excluding examinations with such scan issues, readers con- scans wereof poor quality). Thestudy design and power calcula- sidered DL-CBCTA noninferior to CTA, despite more artifacts. tion aimed to assess the diagnostic quality of the entire intracranial artery vasculature. In the setting of large-vessel occlusion stroke, it may be favorable to power a diagnostic study with regard to num- ACKNOWLEDGMENTS ber of patients or only large-diameter arteries, albeit with the risk We thank all members of the NEXIS consortium for a collaborative of missing information about small vessels contributing to collat- effort. We thank Kevin Najarian for biostatistics support. We thank eral flow and small concurrent occlusions. GavinPoludniowski, ArturOmarand Dirk Schäferfor their support The results in this study are derived by comparing the same in questions regarding medical physics and image reconstruction. vessel segment in the same patient. Consequently, absent vessel segments due to variant anatomy (for example absent A1 or PICA) Disclosure forms provided by the authors are available with the full text and PDF of this article at www.ajnr.org. received the same vessel conspicuity score in both DL-CBCTA and CTA (ie, 1). However, readers were not specifically instructed on REFERENCES how to assess segments with multiple vessels (lenticulostriate 1. Jaffray DA, Siewerdsen JH. Cone-beam computed tomography with arteries, basilar perforating arteries) or certain variant anatomy a flat-panel imager: initial performance characterization. Med Phys (such asan azygosA2 or an AICA-PICA complex). Thisissue may 2000;27:1311–23 CrossRef Medline have increased the interrater variability (k values were only fair). 2. Soderman M, Babic D,Holmin S,et al. Brain imaging with a flat detec- In the majority results, as well as individual reader results of the tor C-arm: Technique and clinical interest of XperCT. Neuroradiology powered data set and subset, we show a significant within-patient 2008;50:863–68 CrossRef Medline 3. Dijkstra ML, Eagleton MJ, Greenberg RK, et al. Intraoperative C- correlation of arterial segments. In the majority results, this is likely arm cone-beam computed tomography in fenestrated/branched influenced by the M1 segment being most frequently rated inferior. aortic endografting. JVasc Surg 2011;53:583–90 CrossRef Medline For individual readers, the lenticulostriate arteries and basilar per- 4. Tacher V, Radaelli A, Lin M, et al. How I do it: cone-beam CT dur- forators are likely contributors in addition to M1 because they ing transarterial chemoembolization for liver cancer. Radiology were interpreted inconsistently. 2015;274:320–34 CrossRef Medline 528 Ståhl May 2023 www.ajnr.org 5. Doerfler A, Golitz P, Engelhorn T, et al. Flat-panel computed to- 16. Keselman B, Berglund A, Ahmed N, et al. The Stockholm Stroke mography (DYNA-CT) in neuroradiology: from high-resolution Triage Project: outcomes of endovascular thrombectomy before imaging of implants to one-stop-shopping for acute stroke. Clin and after triage implementation. Stroke 2022;53:473–81 CrossRef Neuroradiol 2015;25 Suppl 2:291–97 CrossRef Medline Medline 6. Brehm A, Tsogkas I, Maier IL, et al. One-stop management with 17. Brown KM, Zabic S, Shechter G. Impact of spectral separation in perfusion for transfer patients with stroke due to a large-vessel dual-energy CT with anti-correlated statistical reconstruction. occlusion: feasibility and effects on in-hospital times. AJNR J Proceedings of the Fully Three-Dimensional Image Reconstruction in Neuroradiol 2019;40:1330–34 CrossRef Medline Radiology and Nuclear Medicine, Worcester, Massachusetts. July 2, 7. Struffert T, Deuerling-Zheng Y, Kloska S, et al. Flat detector CT in 2015:491–94 the evaluation of brain parenchyma, intracranial vasculature, and 18. Zopfs D, Lennartz S, Abdullayev N, et al. Generally applicable win- cerebral blood volume: a pilot study in patients with acute symp- dow settings of low-keV virtual monoenergetic reconstructions in toms of cerebral ischemia. AJNR Am J Neuroradiol 2010;31:1462–69 dual-layer CT-angiography of the head and neck. Quant Imaging CrossRef Medline Med Surg 2021;11:3408–17 CrossRef Medline 8. Struffert T, Deuerling-Zheng Y, Kloska S, et al. Dynamic angiography 19. Hinkmann FM, Voit HL, Anders K, et al. Ultra-fast carotid CT-an- and perfusion imaging using flat detector CT in the angiography giography: low versus standard volume contrast material protocol suite: a pilot study in patients with acute middle cerebral artery for a 128-slice CT-system. Invest Radiol 2009;44:257–64 CrossRef occlusions. AJNR Am J Neuroradiol 2015;36:1964–70 CrossRef Medline Medline 9. Saake M, Breuer L, Goelitz P, et al. Flat detector computed tomogra- 20. Mantel N, Haenszel W. Statistical aspects of the analysis of data phy angiography with intravenous contrast application: feasibility from retrospective studies of disease. JNatl Cancer Inst 1959;22:719– for visualization of cerebral arterial vasculature. JNeuroimaging 48 Medline 2013;23:414–20 CrossRef Medline 21. Fleiss JL. Measuring nominal scale agreement among many raters. 10. Hoelter P, Goelitz P, Lang S, et al. Visualization of large vessel occlu- sion, clot extent, and collateral supply using volume perfusion flat Psychol Bull 1971;76:378–82 CrossRef detector computed tomography in acute patients with stroke. Acta 22. Cancelliere NM, Hummel E, van Nijnatten F, et al. The butterfly Radiol 2019;60:1504–11 CrossRef Medline effect: improving brain cone-beam CT image artifacts for stroke 11. Schneider D,Apfaltrer P,Sudarski S,et al. Optimization of kiloelec- assessment using a novel dual-axis trajectory. J Neurointerv Surg tron volt settings in cerebral and cervical dual-energy CT angiogra- 2023;15;283–87 CrossRef Medline phy determined with virtual monoenergetic imaging. Acad Radiol 23. Leyhe JR, Tsogkas I, Hesse AC, et al. Latest generation of flat detec- 2014;21:431–36 CrossRef Medline tor CT as a peri-interventional diagnostic tool: a comparative 12. Leithner D, Mahmoudi S, Wichmann JL, et al. Evaluation of virtual study with multidetector CT. J Neurointerv Surg 2017;9:1253–57 monoenergetic imaging algorithms for dual-energy carotid and in- CrossRef Medline tracerebral CT angiography: effects on image quality, artefacts and 24. Ouadah S, Jacobson M, Stayman JW, et al. Correction of patient diagnostic performance for the detection of stenosis. Eur J Radiol motion in cone-beam CT using 3D–2D registration. Phys Med Biol 2018;99:111–17 CrossRef Medline 2017;62:8813–31 CrossRef Medline 13. Neuhaus V, Große Hokamp N, Abdullayev N, et al. Comparison of 25. Li Y, Garrett JW, Li K, et al. Time-resolved C-arm cone beam CT virtual monoenergetic and polyenergetic images reconstructed angiography (TR-CBCTA) imaging from a single short-scan C- from dual-layer detector CT angiography of the head and neck. arm cone beam CT acquisition with intra-arterial contrast injec- Eur Radiol 2018;28:1102–10 CrossRef Medline tion. Phys Med Biol 2018;63:075001 CrossRef Medline 14. Fredenberg E. Spectral and dual-energy X-ray imaging for medical 26. American Association of Physicists in Medicine. Report of American applications. Nuclear Instruments and Methods in Physics Research Association of Physicists in Medicine Task Group 111: The Future Section A: Accelerators, Spectrometers, Detectors and Associated of CT Dosimetry: Comprehensive Methodology for the Evaluation Equipment 2018;878:74–87 CrossRef of Radiation Dose in X-Ray Computed Tomography (2010). 15. Ståhl F, Schäfer D, Omar A, et al. Performance characterization of a https://www.aapm.org/pubs/reports/detail.asp?docid=109.Accessed prototype dual-layer cone-beam computed tomography system. Med Phys 2021;48:6740–54 CrossRef Medline September 6, 2022 AJNR Am J Neuroradiol 44:523–29 May 2023 www.ajnr.org 529 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png American Journal of Neuroradiology American Journal of Neuroradiology

Dual-Layer Detector Cone-Beam CT Angiography for Stroke Assessment: First-in-Human Results (the Next Generation X-ray Imaging System Trial)

Download PDF
8 pages

Loading next page...
 
/lp/american-journal-of-neuroradiology/dual-layer-detector-cone-beam-ct-angiography-for-stroke-assessment-N5aLCEbphS

References (26)

Publisher
American Journal of Neuroradiology
Copyright
© 2023 by American Journal of Neuroradiology. Indicates open access to non-subscribers at www.ajnr.org
ISSN
0195-6108
eISSN
1936-959X
DOI
10.3174/ajnr.a7835
Publisher site
See Article on Publisher Site

Abstract

Dual-Layer Detector Cone-Beam CT Angiography for Stroke Assessment: First-in-Human Results (the Next Generation X-ray Imaging System Trial) F. Ståhl, H. Almqvist, J. Kolloch, Å. Aspelin, V. Gontu, E. Hummel, M. van Vlimmeren, M. Simon, A. Thran, Å. This information is current as Holmberg, M.V. Mazya, M. Söderman and A.F. Delgado of June 3, 2023. AJNR Am J Neuroradiol 2023, 44 (5) 523-529 doi: https://doi.org/10.3174/ajnr.A7835 http://www.ajnr.org/content/44/5/523 ORIGINAL RESEARCH ADULT BRAIN Dual-Layer Detector Cone-Beam CT Angiography for Stroke Assessment: First-in-Human Results (the Next Generation X-ray Imaging System Trial) F. Ståhl, H. Almqvist, J. Kolloch, Å. Aspelin, V. Gontu, E. Hummel, M. van Vlimmeren, M. Simon, A. Thran, Å. Holmberg, M.V. Mazya, M. Söderman, and A.F. Delgado ABSTRACT BACKGROUND AND PURPOSE: In patients with stroke, IV cone-beam CTA in the angiography suite could be an alternative to CTA to shorten the door-to-thrombectomy time. However, image quality in cone-beam CTA is typically limited by artifacts. This study evaluated a prototype dual-layer detector cone-beam CT angiography versus CTA in patients with stroke. MATERIALS AND METHODS: A prospective, single-center trial enrolled consecutive patients with ischemic or hemorrhagic stroke on initial CT. Intracranial arterial segment vessel conspicuity and artifact presence were evaluated on dual-layer cone-beam CTA 70-keV virtual monoenergetic images and CTA. Eleven predetermined vessel segments were matched for every patient. Twelve patients were necessary to show noninferiority to CTA. Noninferiority was determined by the exact binomial test; the 1-sided lower performance boundary was prospectively set to 80% (98.75% CI). RESULTS: Twenty-one patients had matched image sets (mean age, 72 years). After excluding examinations with movement or con- trast media injection issues, all readers individually considered dual-layer cone-beam CT angiography noninferior to CTA (CI bound- ary, 93%, 84%, 80%, respectively) when evaluating arteries relevant in candidates for intracranial thrombectomy. Artifacts were more prevalent compared with CTA. The majority assessment rated each individual segment except M1 as having noninferior con- spicuity compared with CTA. CONCLUSIONS: In a single-center stroke setting, dual-layer detector cone-beam CTA virtual monoenergetic images are noninferior to CTA under certain conditions. Notably, the prototype is hampered by a long scan time and is not capable of contrast media bolus tracking. After excluding examinations with such scan issues, readers considered dual-layer detector cone-beam CTA noninfe- rior to CTA, despite more artifacts. ABBREVIATIONS: CBCT ¼ cone-beam CT; CBCTA ¼ cone-beam CTA; CNR ¼ contrast-to-noise ratio; DL ¼ dual-layer; VMI ¼ virtual monoenergetic images one-beam CT (CBCT) with flat image detectors was first (CBCTA) in the angiography suite could be an alternative to 1 6 Cdemonstrated in 2000. The technique is widely used in CTA to shorten the door-to-thrombectomy time. CBCTA has interventional radiology procedures for anatomic and pathology been shown to be equal or better than CTA for the diagnosis of 7,8 2-5 intracranial stenosis or proximal MCA occlusions. Astudy in assessment. In patients with stroke, IV CBCT angiography 10 patients indicated superior image-quality results in several CBCTA arterial segments compared with CTA; however, the Received October 13, 2022; accepted after revision February 27, 2023. patient cohort was heterogeneous and biased, and the results From the Departments of Neuroradiology (F.S., H.A., J.K., Å.A., V.G., Å.H., M. Söderman, A.F.D.) and Neurology (M.V.M.), Karolinska University Hospital, Stockholm, Sweden; were only presented for individual vessel segments. Aretrospec- Department of Clinical Neuroscience (F.S., H.A., V.G., M.V.M., M. Söderman, A.F.D.) tive study on 16 patients with stroke with large-vessel occlusion Karolinska Institutet, Stockholm, Sweden; Image Guided Therapy (E.H., M.v.V.), Phillips Healthcare, Best, the Netherlands; and Philips Research Hamburg (M. Simon, A.T.), of the anterior circulation indicated that CBCTA generated from Hamburg, Germany. a volume perfusion scan can reliably identify the site of occlusion The trial was supported by a grant from the European Commission (Horizon 2020, in the ICA and M1 segment of the MCA. However, this study NEXIS-project, grant number 780026). did not find CBCTA superior to CTA. Please address correspondence to Fredrik Ståhl, MD, Department of Neuroradiology, Karolinska University Hospital, Eugeniavaegen 3, 171 76 Stockholm, Sweden; e-mail: In CTA, virtual monoenergetic images (VMIs) reconstructed Fredrik.stahl@regionstockholm.se from dual-energy scans have superior image quality and improve Indicates open access to non-subscribers at www.ajnr.org the diagnostic assessment of intracranial vessels compared with 11-13 Indicates article with online supplemental data. conventional polyenergetic images. CBCTA image quality would http://dx.doi.org/10.3174/ajnr.A7835 likely benefit from dual-energy VMI reconstructions by means of AJNR Am J Neuroradiol 44:523–29 May 2023 www.ajnr.org 523 reduced beam-hardening artifacts and an improved contrast-to- the dual-layer detector captures more photons than a conventional 11-14 noise ratio (CNR). Recently, performance characteristics of a CBCT detector. DL-CBCTA VMIs were derived from Compton and photoelectric base projections (basis material decomposition prototype dual-layer detector CBCT (DL-CBCT) system were was performed in the projection domain). Vessel conspicuity may published, with the possibility of optimizing CNR by VMI energy be improved due to intrinsic compensation of iodine beam-harden- selection. ing artifacts in DL-CBCTA VMIs. Anticorrelated noise reduction Because previous studies on CBCTA are small, heterogeneous, was exploited as an inherent benefit of DL-CBCT. All DL-CBCTA and selective, a trial enrolling consecutive patients is necessary to and CTA scans used the same automated injection protocol: 85 mL determine the role of CBCTA in the primary diagnostic setting of 7-10 of iodine, 320 mg I/mL (iodixanol, Visipaque; GE Healthcare), suspected stroke. In this prospective trial, we assessed arterial 5 mL/s into a peripheral vein followed by an 80-mL saline chaser. visibility and artifact presence in patients with stroke when exam- The second-phase CTA standard reconstruction was included ined with DL-CBCT angiography (DL-CBCTA) and CTA. To our in the study as a reference standard because it was reconstructed knowledge, this is the first trial on CBCTA with prospectively with a smaller FOV and the timing was typically more similar to defined end points, enrolling awake patients with the aim of inves- the DL-CBCTA contrast phase (late arterial/arteriovenous). tigating noninferiority to CTA. Furthermore, this work presents Phantom measurements indicated that the DL-CBCTA and CTA the first-in-human results from a DL-CBCT prototype system. systems had comparable high-contrast spatial resolution. Scan details are shown in Table 1. In conjunction with the CTA and MATERIALS AND METHODS DL-CBCTA, the quality of the image acquisition was evaluated Trial Design and Participants with regard to patient motion and timing of the contrast media The Next Generation X-ray Imaging System (NEXIS) trial was a injection. Scans were rated as good, acceptable, or poor. prospective, nonrandomized open-label single-sequence, 2-period crossover clinical trial with blinded image readers conducted at a Diagnostic Image-Quality Assessment comprehensive stroke center. The study sought to include patients A series of pilot studies were performed to determine the optimal evaluated for endovascular thrombectomy on arrival at the hospi- VMI energy and noise-reduction levels for the DL-CBCTA tal. Patients with a suspected large-vessel occlusion of the anterior images. We randomized 50- to 80-keV VMIs at 3 different noise- circulation underwent prehospital triage to bypass primary stroke 16 reduction levels, and expert readers evaluated them side-by-side centers as per clinical routine. Patients 50 years of age or older blinded to noise and energy levels (F.S., H.A., V.G., M. Söderman, with ischemic stroke of the anterior circulation or hemorrhagic A.F.D.). Window settings were normalized to a previous publica- stroke were consecutively enrolled from November 2020 to April tion on dual-energy CTA. Images were ranked in order of pref- 2021. Initial imaging was with CTA as per clinical routine and sub- erence with regard to image quality and artifact presence. The 70- sequently with the prototype DL-CBCT in an adjacent room. keV images with a moderate noise reduction were best suited for Depending on applicable eligibility criteria, patients were imaged the purpose of the study. once or twice with DL-CBCT (the same day as CTA or/and 1 day af- In the reader study, vessel conspicuity and artifact presence ter CTA). Eligibility criteria, inclusion groups, and flow diagrams of were evaluated separately on 5-point Likert scales, adopted with study participation arepresented in theOnlineSupplemental Data. slight modifications from previous studies (5, excellent vessel con- The study protocol and informed consent forms were approved spicuity or no artifacts; 1, vessel not visible or extensive artifacts) by the Swedish Ethical Review Authority (approval No. 2020– 10,19 (see the Online Supplemental Data for a detailed description). 00157). The prototype DL-CBCT system was approved by the All patients had unilateral ischemic or hemorrhagic stroke; thus, Swedish Medical Products Agency (document No. 5.1–2020– the affected hemispheres were not included in the analysis. Images 6325), in accordance with the Medical Device Directive. Philips were randomized and evaluated independently by 3 neuroradiolo- Healthcare was the formal sponsor of the study because of direct gists (H.A., J.K., and Å.A.) each with .9 years of experience in a liability for the prototype DL-CBCT system. An independent quali- single-sequence, 2-period crossover design with a mean washout fied research organization contracted by the sponsor monitored the period of 4 weeks (range, 1–8 weeks). The readers were blinded to study. All patients signed an informed consent. The trial was sup- technique, and there was no clinical information. The study soft- ported by a grant from the European Commission (Horizon 2020, ware allowed changes in section thickness, viewing plane, window NEXIS-project, grant No. 780026). The study was registered pro- level, and window width. spectively at clinicaltrials.gov (identifier: NCT04571099). Sixteen intracranial arterial segments were prospectively defined for the readers’ study. For the powered analysis, some CTA and Prototype DL-CBCTA arterial segments were merged to render 11 arterial segments The systems used for CTA were Aquilion ONE (Canon Medical Systems) and IQon Spectral CT (Philips Healthcare). The examina- per patient (Table 2). For merged segments, the scores of indi- tion was according to clinical routine, typically including noncon- vidually evaluated segments were averaged. The score difference trast CT of the head, multiphase CTA, and CTP. The prototype between DL-CBCTA and CTA for each segment in each patient DL-CBCT system (Allura NEXIS Investigational Device; Philips determined whether a segment was considered inferior, equal, Healthcare) was a commercial interventional C-arm x-ray system or superior. In case of an absent segment due to variant anat- omy, a vessel conspicuity score of 1 was given for the specific (Allura Xper FD20/15; Philips Healthcare) fitted with a dual-layer segment (see the previous paragraph). In addition to individual 20-inch (379.4  293.2 mm) non-CE marked detector prototype. The prototype detector has been previously described. In essence, reader’s results, the majority assessment is also presented, ie, 524 Ståhl May 2023 www.ajnr.org Table 1: Scan details Canon Aquilion ONE Philips IQon Prototype DL-CBCT Tube (kV) 100 120 120 Tube current (mAs/mA) (average) 196 (Auto-modulation) 114 (Auto-modulation) 310 Scan time/rotation time (sec) 3.3/0.5 (Full rotation) 2.5/0.3 (Full rotation) 20.0/20.0 (200° rotation) Nominal beam width (mm) 80  0.500 64  0.625 194.700 Pitch factor 0.813 0.671 NA Display FOV coronal  sagittal  axial (mm ) 210.9  210.9  160.0 210.0  210.0  160.0 251.8  251.8  194.7 Section thickness (mm) 0.50 0.67 0.66 Matrix size 512  512 512  512 384  384 15a Reconstruction kernel FC43 Filter UA Ståhl et al 4 15a Reconstruction algorithm AIDR 3D eStandard iDose level 4 Ståhl et al Avg CTDIvol (16-cm phantom) 20.0 mGy 21.3 mGy NA Air kerma (in an 18-cm water phantom) NA NA 57.6 mGy MTF 50%: 3.78 50%: 3.46 50%: 3.57 10%: 6.57 10%: 6.65 10%: 6.04 Note:—CTDIvol indicates volume CT dose index; MTF, modulation transfer function; NA, not applicable. a 15 Details of the prototype algorithm are described in Ståhl et al. Air kerma in an 18-cm diameter plastic water phantom at the center of the scan length, measured in accordance with American Association of Physicists in Medicine Task Group Report 111. Generated from consecutive scans of the upper bead of a Catphan CTP528 module (The Phantom Laboratory). Table 2: Arterial segments RESULTS 11 Segments 11 Segments Population Characteristics 16 Segments (Powered) (Thrombectomy) Of 28 consecutively enrolled patients, 5 had no in-house CTA ICA ICA ICA and 2 underwent subcutaneous IV contrast media injection dur- M1 M1 M1 ing the DL-CBCTA scan. Two patients were imaged twice with M2 M2 M2 DL-CBCTA, and for those patients, the results from both scans M3 M3-M4 M3 were averaged. Consequently, 21 complete and matched DL- M4 M4 A1 A1-A2 A1 CBCTA and CTA image sets from 21 patients were included A2 A2 (Fig 1, flow diagram). The mean age was 72 (SD, 9) years, and Lenticulostriate Lenticulostriate 14 were women (67%). The right hemisphere was affected in 11 Vertebral Vertebral Vertebral (57%) patients. Three patients had a hemorrhagic stroke. Of the Basilar Basilar Basilar AICA AICA-PICA-SCA remaining 18 patients with ischemic stroke, 17 (94%) presented PICA with occlusion of the ICA or proximal MCA (M1 or M2 seg- SCA ment) on CTA. Seventeen patients (81%) were imaged with DL- Basilar perforators Basilar perforators CBCTA the day after CTA imaging (mean, 23.8 [SD, 3.0] hours), P1 P1-P2 P1 of whom 15 had been treated with thrombectomy the day before P2 P2 (modified TICI grade 2C reperfusion or better in 11 [73%]). Note:—Lenticulostriate indicates lenticulostriate artery perforators; Vertebral, in- tracranial vertebral artery; Basilar perforators, basilar artery perforating branches; Because only the unaffected hemisphere was included in the sta- SCA, superior cerebellar artery. tistical analysis, the degree of reperfusion was not considered to influence the results. Four patients (19%) were only imaged segments judged as superior/equal or inferior compared to CTA with DL-CBCTA on the same day as CTA (mean, 1.2 [SD, 0.6] by at least 2 out of 3 readers. hours after CTA). Statistical Analysis The powered outcome of this study was the proportion of arterial Vessel Visibility Assessment segments with equal or superior visibility compared with CTA For the powered end point of overall arterial visibility, 21 patients (Table 2). The target proportion was 90% equal or superior ratings, with a total of 231 matched arterial segments (11 per patient) with a lower boundary performance goal set to 80%. The 1-sided a were evaluated by each reader. One reader scored equal or supe- was modified from .025 to .0125, to account for multiple end rior vessel visibility for DL-CBCTA and CTA in 90% of arterial points. A sample size of 126 was required to render 80% power. segments (CI lower boundary of 84%) and met the predefined Because the powered end point was trait-based, 12 patients with a total of 132 arterial segments (11 segments each) constituted the noninferiority criteria (lower performance boundary of .80%), minimum sample size. For these analyses, the exact binomial test whereas the other 2 readers did not (CI lower performance with a 1-sided 98.75% CI was used. Within-subject correlation of boundary of 58% and 53%, respectively). The result of the major- arterial segments was assessed by the Cochran-Mantel-Haenszel ity analysis was 77% of arterial segments being rated noninferior 20 21 test. Interrater agreement was assessed by the Fleiss k. in DL-CBCTA (CI lower boundary of 71%). The Fleiss k between readers was 0.25 (fair agreement). Examples of patient scans with Statistical analyses and graphic presentations were made in R no disagreement among readers and with considerable interrater Studio (Version 1.4.1103; http://rstudio.org/download/desktop). A biostatistician was consulted at all stages of the study. variability are presented in the Online Supplemental Data. AJNR Am J Neuroradiol 44:523–29 May 2023 www.ajnr.org 525 98%, 91%, and 88%, and all readers individually exceeded the CI Post Hoc Subset Analyses lower performance boundary of 80% (Table 3). Majority results Nine of the 21 (43%) DL-CBCTA scans were of poor quality due were identical to those of the powered subset (98%, CI lower bound- to patient motion or suboptimal timing of the contrast media ary 93%). The Fleiss k for this data set was 0.38 (fair agreement). injection (Fig 2). No CTA scan was rated as poor. In a subset There was no significant within-patient correlation of individual ar- analysis including only the 12 matched scans of acceptable or good quality, the 3 readers rated vessel visibility as noninferior terial segments for any of the individual readers in the thrombec- in 98%, 88%, and 78% of the cases, respectively, which meant tomy data set or subset. However, a significant correlation was that 2 readers met the predefined performance goal (CI lower found for all majority analyses and most individual reader results boundary of .80%). The majority result for this subset was 98% for the powered data set and subset (Online Supplemental Data). (CI lower boundary of 93%). See Table 3 for a detailed display of the results. Image Artifact Assessment Eleven arterial segments especially relevant for thrombectomy Image artifacts were assessed in the same arterial segments as used candidates were selected for evaluation in the same patient subset for vessel visibility. For the prospectively determined arterial seg- (Table 2). According to each reader, noninferior vessel visibility was ments, readers perceived artifact presence as noninferior to CTA in 58%, 42%, and 36% of the segments rated (Table 4). The Fleiss k was 0.26 (fair agreement). After excluding the 9 scans of inferior quality, noninferiority to CTA for artifact assessment was per- ceived in 68%, 66%, and 55% of the cases. The lowest number of artifacts compared with CTA was observed in the subset of 11 arterial segments espe- cially relevant for thrombectomy candi- dates (Table 4). The Fleiss k for this data set was 0.34 (fair agreement). In summary, noninferiority in terms of artifacts was not seen in the full data set or any subset analysis, given a lower performance boundary of 80%. Evaluation of Individual Arterial FIG 1. Of 28 consecutively enrolled patients, 5 had no in-house CTA and 2 had subcutaneous IV Segments contrast media injection during the DL-CBCTA scan. Two patients were imaged twice with DL- CBCTA, and for those, the results from both scans were averaged. Consequently, 21 complete The Online Supplemental Data show and matched DL-CBCTA and CTA image sets from 21 patients were included. the whole data set of individual arterial FIG 2. DL-CBCTA 70-keV images (upper row)and CTA (lower row) with MIP 35-mm section thickness. A and B, Acceptable-quality DL-CBCTA scans. C and D, Typical scans in the data set affected by motion artifacts. Lower row (E–H) shows the corresponding CTA. Note that images F, G, and H show a right-sided MCA occlusion, which have been resolved at the time for DL-CBCTA imaging. Only the arterial anatomy in the unaf- fected hemisphere was evaluated in this study. 526 Ståhl May 2023 www.ajnr.org a Table 3: Vessel visibility DISCUSSION Powered Data Powered Thrombectomy Data Thrombectomy Thestudy compared theintracranial ar- Set Subset Set Subset tery visibility of IV DL- CBCTA images Patients 21 12 21 12 with the reference standard CTA in con- Segments 231 132 231 132 secutively enrolled patients with stroke. rated b b Of dual-detector CBCTA scans, 43% Majority 0.77 (0.70) 0.98 (0.93) 0.77 (0.71) 0.98 (0.93) b b were of poor quality due to movement Reader 1 0.65 (0.58) 0.88 (0.80) 0.67 (0.60) 0.91 (0.84) b b b b Reader 2 0.90 (0.84) 0.98 (0.93) 0.89 (0.83) 0.98 (0.93) artifacts or contrast media injection Reader 3 0.60 (0.53) 0.78 (0.69) 0.68 (0.61) 0.88 (0.80) issues. In the subset of patients with ac- Proportion of DL-CBCTA arterial segment visibility rated equal or superior to CTA. The data set (21 patients) ceptable scan quality, DL-CBCTA was includes all scans; the subset (12 patients) excluded inferior scans. The 98.75% CI of the 1-sided lower performance noninferior to CTA with regard to intra- boundary is in parentheses (lower boundary is defined as 80% rated equal or superior). Statistically significant result. cranial artery conspicuity, despite more prominent image artifacts. The strengths of the trial were the Table 4: Artifacts prospectively defined end points, the Powered Data Powered Thrombectomy Data Thrombectomy clinically relevant study population, Set Subset Set Subset and the assessment by blinded readers. Patients 21 12 21 12 Segments 231 132 231 132 The noninferiority lower boundary was rated set to 80% to minimize the risk of type Majority 0.41 (0.34) 0.63 (0.53) 0.55 (0.48) 0.81 (0.72) I error. For the complete data set, 1 Reader 1 0.42 (0.35) 0.68 (0.58) 0.54 (0.46) 0.85 (0.77) reader determined DL-CBCTA arterial Reader 2 0.58 (0.50) 0.66 (0.56) 0.65 (0.58) 0.74 (0.64) visibility as noninferior to CTA, but the Reader 3 0.36 (0.29) 0.55 (0.44) 0.52 (0.44) 0.73 (0.63) a other 2 readers did not. However, after Proportion of DL-CBCTA arterial segment artifacts rated equal or superior to CTA. Data set (21 patients) includes all scans; subset (12 patients) excluded inferior scans. The 98.75% CI of the 1-sided lower performance boundary is excluding DL-CBCTA scans with in parentheses (lower boundary is defined as 80% rated equal or superior). movement artifacts or suboptimal con- trast media injections, 2 readers indi- vidually agreed on noninferiority of DL-CBCTA versus CTA. In this subset, the majority score deter- segment ratings per reader and according to the majority assess- mined vessel conspicuity to be noninferior to CTA. Image arti- ment. In summary,A2-,M4- and AICA-segment vesselcon- facts were generally more abundant in the DL-CBCTA images, spicuity was most commonly rated as noninferior to CTA (at and the results did not reach noninferiority. least 76% for each reader and segment), with high majority After we excluded scans of poor quality, all arterial segments scores (range, 81%–90% rated noninferior). The lenticulostri- except M1 showed noninferior conspicuity to CTA according to ate and basilar perforating arteries had high majority scores the majority assessment. With regard to individual reader results, for vessel conspicuity; however, 1 reader’s results largely dif- theM4, A2, P2, and AICA segments had thebestvessel visibility in fered from the other 2. It was noted that a safe distinction the subset, and all except AICA expectedly indicated a low degree from small draining veins was difficult. Consequently, the ma- of artifacts in the corresponding segments. The M1 and ICA seg- jority score was interpreted with caution for these segments. ments had the lowest scores for vessel visibility compared with The A2 segment was rated with the fewest artifacts (majority CTA and indicated a higher prevalence of artifacts. Our results score, 76% rated noninferior), followed by A1, M2, M3, M4, indicate that arteries that are not influenced by skull base artifacts P1, and P2 segments (majority score, 62%–67%, were rated may have superior conspicuity on DL-CBCTA compared with noninferior). CTA. As the systems had comparable spatial resolution, this may In the subset of 12 acceptable scans (Online Supplemental be attributed to fundamental differences of the VMI (DL-CBCTA) Data), A2, M4, AICA, and P2 segments were the most commonly and polyenergetic (CTA) reconstructions. DL-CBCTA VMIs had a rated as noninferior vessel conspicuity to CTA (at least 92% for lower absolute noise (achieved through anticorrelated noise reduc- each reader and segment). According to the majority assessment, tion) and may be superior in mitigating beam-hardening artifacts 100% noninferior vessel conspicuity to CTA was seen for all seg- from iodinated vessels (due to material decomposition in the pro- ments except M1 (75% rated noninferior). For artifacts, 100% of jection space) compared with CTA. In contrast, arteries adjacent to A1, A2, M3, M4, and P2 segments were noninferior in the subset the skull base showed an inferior visibility compared with CTA. according to the majority assessment. Even though beam-hardening artifacts are mitigated by VMIs, the M1-segment vessel conspicuity was noninferior to CTA in findings were expected because circular CBCT scans are highly 57% of the whole data set and 75% of the subset, the lowest susceptible to skull base artifacts such as beam-hardening. Since scores registered. The M1 segment also had a higher degree of the M1 and ICA segments are often involved in acute ischemic artifacts (43% rated noninferior in the data set and 67% in the stroke, future effort should focus on improving visualization of subset). The intracranial ICA segment similarly had a high these segments. Recently, a novel x-ray tube trajectory for CBCT prevalence of artifacts, and its scores were among the lowest was shown to improve intracranial image quality by reducing bone registered. beam-hardening artifacts from the skull. AJNR Am J Neuroradiol 44:523–29 May 2023 www.ajnr.org 527 The DL-CBCTA scan time was 20 seconds to enable dual- A minority of patients (19%) were imaged with DL-CBCTA layer image acquisition. Nine scans (43%) were considered of shortly after CTA, which may have affected the distribution of io- poor quality, predominantly due to motion artifacts. No repeat dine contrast media in the DL-CBCTA images. The impact of scans were performed in this study, in accordance with the potential differences in flow dynamics and vessel status for patients ethics approval and study protocol. Most interesting, all except imaged with DL-CBCTA 1 day after CTA is unknown. CNR com- 1 DL-CBCTA scan obtained on the day of admission showed parison of DL-CBCTA and CTA was not possible because image some degree of motion artifacts despite efforts to fixate the head acquisition timing after IV contrast injections were not identical. A in the head rest. The long scan time is a characteristic disadvant- previous study on CBCTA used a 512  512 image pixel matrix, age of CBCT, which must be considered when selecting an similar to CT. Using a 384  384 image pixel matrix for DL- appropriate imaging technique for patients with stroke. In CBCTA impacts the level of detail and noise characteristics, possi- bly affecting the perception of image quality. The DL-CBCTA scan awake patients with limited compliance, efforts to minimize the time of 20 seconds made it prone to movement artifacts. Also, in risk of patient movement is warranted. An enhanced head fixa- some cases, image quality was poor due to suboptimal timing of tion may decrease motion artifacts. Moreover, novel recon- the contrast media injection. Because the trial studied a prototype struction methods may mitigate the effects of patient motion in 24,25 system,manualbolus tracking wasnot possible.Instead,a manual CBCT. If there are still considerable motion artifacts pres- delay of 15–20 seconds was used for all patients. Furthermore, we ent, one may consider a repeat scan. used only 1 acquisition protocol. Future studies should involve Earlier studies evaluating the diagnostic accuracy of CBCTA image acquisition and reconstruction protocol optimization. either had patients under general anesthesia or a highly selective DL-CBCTA after IV contrast media injection in angiography patient cohort able to comply with instructions. As expected, suite enables the primary diagnostic work-up and subsequent treat- these studies did not report any significant impact on image qual- ment in the same room. In the clinical setting, this approach would ity from motion artifacts. In 1 study, it is unclear whether analy- require concurrent image evaluation of the brain to rule out intra- ses were matched pair-wise, and the results were presented only cranial hemorrhage and to enable assessment of irreversible brain is- for individual vessel segments and not for the CBCTA as a chemia and possibly brain perfusion. Noncontrast DL-CBCT of the whole. In addition, outcome measures from previous studies brain has been studied within the scope of the NEXIS trial, and data were not prospectively defined and typically were not adjusted 7-10 are currently being prepared for final analysis. Visualization of the for multiple comparisons. cervical vasculature is warranted in patients eligible for thrombec- This trial has several limitations and technical considerations. tomy; however, this was not addressed in the scope of this study. The ethics approval did not allow inclusion of patients before treat- ment. Thus, nonsymptomatic vessel segment conspicuity and arti- facts were evaluated as characteristics of diagnostic quality. The CONCLUSIONS original intention was to perform a subset analysis on the diagnos- In a single-center stroke setting, DL-CBCTA 70-keV virtual tic accuracy to identify intracranial occlusions in a subgroup of monoenergetic images are noninferior to CTA under certain con- patients (inclusion group 3.I, Online Supplemental Data), but the ditions. Notably, the prototype system is hampered by a long analysis was not performed due to few patients (n ¼ 3, of which 2 scan time and is not capable of contrast media bolus tracking. After excluding examinations with such scan issues, readers con- scans wereof poor quality). Thestudy design and power calcula- sidered DL-CBCTA noninferior to CTA, despite more artifacts. tion aimed to assess the diagnostic quality of the entire intracranial artery vasculature. In the setting of large-vessel occlusion stroke, it may be favorable to power a diagnostic study with regard to num- ACKNOWLEDGMENTS ber of patients or only large-diameter arteries, albeit with the risk We thank all members of the NEXIS consortium for a collaborative of missing information about small vessels contributing to collat- effort. We thank Kevin Najarian for biostatistics support. We thank eral flow and small concurrent occlusions. GavinPoludniowski, ArturOmarand Dirk Schäferfor their support The results in this study are derived by comparing the same in questions regarding medical physics and image reconstruction. vessel segment in the same patient. Consequently, absent vessel segments due to variant anatomy (for example absent A1 or PICA) Disclosure forms provided by the authors are available with the full text and PDF of this article at www.ajnr.org. received the same vessel conspicuity score in both DL-CBCTA and CTA (ie, 1). However, readers were not specifically instructed on REFERENCES how to assess segments with multiple vessels (lenticulostriate 1. Jaffray DA, Siewerdsen JH. Cone-beam computed tomography with arteries, basilar perforating arteries) or certain variant anatomy a flat-panel imager: initial performance characterization. Med Phys (such asan azygosA2 or an AICA-PICA complex). Thisissue may 2000;27:1311–23 CrossRef Medline have increased the interrater variability (k values were only fair). 2. Soderman M, Babic D,Holmin S,et al. Brain imaging with a flat detec- In the majority results, as well as individual reader results of the tor C-arm: Technique and clinical interest of XperCT. Neuroradiology powered data set and subset, we show a significant within-patient 2008;50:863–68 CrossRef Medline 3. Dijkstra ML, Eagleton MJ, Greenberg RK, et al. Intraoperative C- correlation of arterial segments. In the majority results, this is likely arm cone-beam computed tomography in fenestrated/branched influenced by the M1 segment being most frequently rated inferior. aortic endografting. JVasc Surg 2011;53:583–90 CrossRef Medline For individual readers, the lenticulostriate arteries and basilar per- 4. Tacher V, Radaelli A, Lin M, et al. How I do it: cone-beam CT dur- forators are likely contributors in addition to M1 because they ing transarterial chemoembolization for liver cancer. Radiology were interpreted inconsistently. 2015;274:320–34 CrossRef Medline 528 Ståhl May 2023 www.ajnr.org 5. Doerfler A, Golitz P, Engelhorn T, et al. Flat-panel computed to- 16. Keselman B, Berglund A, Ahmed N, et al. The Stockholm Stroke mography (DYNA-CT) in neuroradiology: from high-resolution Triage Project: outcomes of endovascular thrombectomy before imaging of implants to one-stop-shopping for acute stroke. Clin and after triage implementation. Stroke 2022;53:473–81 CrossRef Neuroradiol 2015;25 Suppl 2:291–97 CrossRef Medline Medline 6. Brehm A, Tsogkas I, Maier IL, et al. One-stop management with 17. Brown KM, Zabic S, Shechter G. Impact of spectral separation in perfusion for transfer patients with stroke due to a large-vessel dual-energy CT with anti-correlated statistical reconstruction. occlusion: feasibility and effects on in-hospital times. AJNR J Proceedings of the Fully Three-Dimensional Image Reconstruction in Neuroradiol 2019;40:1330–34 CrossRef Medline Radiology and Nuclear Medicine, Worcester, Massachusetts. July 2, 7. Struffert T, Deuerling-Zheng Y, Kloska S, et al. Flat detector CT in 2015:491–94 the evaluation of brain parenchyma, intracranial vasculature, and 18. Zopfs D, Lennartz S, Abdullayev N, et al. Generally applicable win- cerebral blood volume: a pilot study in patients with acute symp- dow settings of low-keV virtual monoenergetic reconstructions in toms of cerebral ischemia. AJNR Am J Neuroradiol 2010;31:1462–69 dual-layer CT-angiography of the head and neck. Quant Imaging CrossRef Medline Med Surg 2021;11:3408–17 CrossRef Medline 8. Struffert T, Deuerling-Zheng Y, Kloska S, et al. Dynamic angiography 19. Hinkmann FM, Voit HL, Anders K, et al. Ultra-fast carotid CT-an- and perfusion imaging using flat detector CT in the angiography giography: low versus standard volume contrast material protocol suite: a pilot study in patients with acute middle cerebral artery for a 128-slice CT-system. Invest Radiol 2009;44:257–64 CrossRef occlusions. AJNR Am J Neuroradiol 2015;36:1964–70 CrossRef Medline Medline 9. Saake M, Breuer L, Goelitz P, et al. Flat detector computed tomogra- 20. Mantel N, Haenszel W. Statistical aspects of the analysis of data phy angiography with intravenous contrast application: feasibility from retrospective studies of disease. JNatl Cancer Inst 1959;22:719– for visualization of cerebral arterial vasculature. JNeuroimaging 48 Medline 2013;23:414–20 CrossRef Medline 21. Fleiss JL. Measuring nominal scale agreement among many raters. 10. Hoelter P, Goelitz P, Lang S, et al. Visualization of large vessel occlu- sion, clot extent, and collateral supply using volume perfusion flat Psychol Bull 1971;76:378–82 CrossRef detector computed tomography in acute patients with stroke. Acta 22. Cancelliere NM, Hummel E, van Nijnatten F, et al. The butterfly Radiol 2019;60:1504–11 CrossRef Medline effect: improving brain cone-beam CT image artifacts for stroke 11. Schneider D,Apfaltrer P,Sudarski S,et al. Optimization of kiloelec- assessment using a novel dual-axis trajectory. J Neurointerv Surg tron volt settings in cerebral and cervical dual-energy CT angiogra- 2023;15;283–87 CrossRef Medline phy determined with virtual monoenergetic imaging. Acad Radiol 23. Leyhe JR, Tsogkas I, Hesse AC, et al. Latest generation of flat detec- 2014;21:431–36 CrossRef Medline tor CT as a peri-interventional diagnostic tool: a comparative 12. Leithner D, Mahmoudi S, Wichmann JL, et al. Evaluation of virtual study with multidetector CT. J Neurointerv Surg 2017;9:1253–57 monoenergetic imaging algorithms for dual-energy carotid and in- CrossRef Medline tracerebral CT angiography: effects on image quality, artefacts and 24. Ouadah S, Jacobson M, Stayman JW, et al. Correction of patient diagnostic performance for the detection of stenosis. Eur J Radiol motion in cone-beam CT using 3D–2D registration. Phys Med Biol 2018;99:111–17 CrossRef Medline 2017;62:8813–31 CrossRef Medline 13. Neuhaus V, Große Hokamp N, Abdullayev N, et al. Comparison of 25. Li Y, Garrett JW, Li K, et al. Time-resolved C-arm cone beam CT virtual monoenergetic and polyenergetic images reconstructed angiography (TR-CBCTA) imaging from a single short-scan C- from dual-layer detector CT angiography of the head and neck. arm cone beam CT acquisition with intra-arterial contrast injec- Eur Radiol 2018;28:1102–10 CrossRef Medline tion. Phys Med Biol 2018;63:075001 CrossRef Medline 14. Fredenberg E. Spectral and dual-energy X-ray imaging for medical 26. American Association of Physicists in Medicine. Report of American applications. Nuclear Instruments and Methods in Physics Research Association of Physicists in Medicine Task Group 111: The Future Section A: Accelerators, Spectrometers, Detectors and Associated of CT Dosimetry: Comprehensive Methodology for the Evaluation Equipment 2018;878:74–87 CrossRef of Radiation Dose in X-Ray Computed Tomography (2010). 15. Ståhl F, Schäfer D, Omar A, et al. Performance characterization of a https://www.aapm.org/pubs/reports/detail.asp?docid=109.Accessed prototype dual-layer cone-beam computed tomography system. Med Phys 2021;48:6740–54 CrossRef Medline September 6, 2022 AJNR Am J Neuroradiol 44:523–29 May 2023 www.ajnr.org 529

Journal

American Journal of Neuroradiology – American Journal of Neuroradiology

Published: May 1, 2023

There are no references for this article.